ChanServ changed the topic of #linux-rockchip to: Rockchip development discussion
kevery has joined #linux-rockchip
Tenkawa has quit [Quit: ... ... ... ... ...]
Tenkawa has joined #linux-rockchip
stikonas has quit [Ping timeout: 248 seconds]
Tenkawa has quit [Quit: ... ... ... ... ...]
Tenkawa has joined #linux-rockchip
Tenkawa has quit [Client Quit]
lurchi_ has joined #linux-rockchip
lurchi__ has quit [Ping timeout: 240 seconds]
camus has quit [Read error: Connection reset by peer]
camus has joined #linux-rockchip
camus has quit [Client Quit]
npcomp has quit [Ping timeout: 252 seconds]
macromorgan has quit [Read error: Connection reset by peer]
npcomp has joined #linux-rockchip
Net147 has quit [Quit: Quit]
Net147 has joined #linux-rockchip
Net147 has quit [Changing host]
Net147 has joined #linux-rockchip
indy has joined #linux-rockchip
ldevulder has joined #linux-rockchip
Rathann has joined #linux-rockchip
npcomp has quit [Ping timeout: 240 seconds]
<montjoie> hello I seek TRM for 3588 and pxa30 where crypto manual is present
<montjoie> I found only pxa30 part1 without crypto
matthias_bgg has joined #linux-rockchip
Jmabsd has quit [Remote host closed the connection]
Jmabsd has joined #linux-rockchip
alpernebbi has quit [Ping timeout: 260 seconds]
alpernebbi has joined #linux-rockchip
dliviu has quit [Ping timeout: 246 seconds]
Jmabsd has quit [Changing host]
Jmabsd has joined #linux-rockchip
<montjoie> links given by cnx-software was removed for rk3588
npcomp has joined #linux-rockchip
<Jmabsd> (OT) Guys, can you help me figure, what are the NXP LX2160A's virtual memory capacity supposedly
<Jmabsd> psydroid,mps,maz: ? =)
<Jmabsd> It's an A72, is this hardcoded per ARM core revision
dliviu has joined #linux-rockchip
<maz> Jmabsd: the A72 TRM has the amswers to these questions.
<maz> answers*
<Jmabsd> TRM = ?
<Jmabsd> ahh, technical reference manual
<Jmabsd> maz: Sorry where in the TRM is this said
<Jmabsd> I think there are two separate hardware limits:
<Jmabsd> The first one is a general limit on the virtual memory address space. This limits the addressable memory one OS process can have on the CPU.
<Jmabsd> But then, I think for VM:s, those VM:s may have a size constraint too which then that VM's OS and any OS process inside it, are subject to
<maz> the former is the VA space, the latter is the IPA space. not the same thing.
<maz> and no, the two aren't related, even in a VM.
<Jmabsd> maz: I see. I don't find "VA space" in the TRM
<Jmabsd> maz: how does IPA space affect the hard metal running OS/processes or a VM guest OS/process?
<maz> it's how much physical address space you have. this applies equally to bare metal and guests.
<Jmabsd> maz: and how much is it for the A72?
<mps> could it depend on particular SoC, i.e. how many address line are wired
<maz> no.
<mps> lines*
<maz> you don't need anything wired for these limits to take effect.
<mps> ah, question is about limits
Rathann has quit [Remote host closed the connection]
<Jmabsd> maz,mps: so the LX2160A = an A72 should have how much virtual memory for each OS process
<maz> up to 48 bits.
stikonas has joined #linux-rockchip
<Jmabsd> maz: aha, i see. where did you see that written up? so that means.. sec
<Jmabsd> maz,mps: great so VA space is 48 bits = 256TB. How does the IPA space affect what software/VM:s can do?
<maz> Jmabsd: it turns out I can count. 4 levels of PT, 4kB per page, 9 bits per level -> 48bits.
<maz> as for the IPA size, it constraints the memory map of your VM.
<Jmabsd> cool I see
<maz> but please move this thread somewhere else. this isn't on-topic at all.
stikonas has quit [Remote host closed the connection]
stikonas has joined #linux-rockchip
<Jmabsd> Yea. I thought there used to be a general ARM conversation, on FreeNode, it was called #arm.
<Jmabsd> Then FreeNode became Libera. It's not there now.
<maz> there is #armlinux, which can be used for this.
lurchi_ is now known as lurchi__
mps has quit [Ping timeout: 256 seconds]
mps has joined #linux-rockchip
<Jmabsd> maz: Thanks a lot for the reference. I will direct further ARM queries there.
Tenkawa has joined #linux-rockchip
dliviu has quit [Ping timeout: 246 seconds]
dliviu has joined #linux-rockchip
lurchi__ is now known as lurchi_
stikonas has quit [Ping timeout: 260 seconds]
lurchi_ is now known as lurchi__
Whistler has quit [Quit: Connection closed for inactivity]
LinuxHackerman[m has joined #linux-rockchip
LinuxHackerman has left #linux-rockchip [#linux-rockchip]
<montjoie> smaeul: thanks
SallyAhaj has quit [Remote host closed the connection]
SallyAhaj has joined #linux-rockchip
SallyAhaj has quit [Remote host closed the connection]
SallyAhaj has joined #linux-rockchip
<maz> oh, MMU600!
lurchi__ is now known as lurchi_
<robmur01> maz: Indeed! Book your seats now for the VFIO vs. non-coherent PCI title fight
<robmur01> (in about a year once I've finished fixing the IOMMU API to allow both rockchip-iommu and arm-smmu-v3 to coexist...)
<maz> robmur01: non-cohe... oh crap...
SallyAhaj has quit [Remote host closed the connection]
SallyAhaj has joined #linux-rockchip
<robmur01> maz: gotta match the still-non-coherent GIC - https://github.com/radxa/kernel/commit/0b985f29304dcb9d644174edacb67298e8049d4f
<maz> robmur01: "here's a nickel, kid. go buy yourself a real interconnect..."
<maz> i also have to laugh at the CONFIG_NO_GKI... how do they do it with GKI? drop the ITS from the DT?
<robmur01> easy, just don't support PCI on Android :)
<maz> interestingly, the TRM vaguely implies that there are *two* ITSs. yet the DT has only one.
<maz> maybe downstream of the other SMMU?
<maz> that'd explain why they didn't use the monolithic configuration...
<robmur01> looks like one for the PCIe3.0 RCs and the other for the PCIe2.0 ones
<maz> ah, right. not looking at the right version, i guess.
* robmur01 gets fed up of Github navigation/search and pulls the branch...
<maz> I expect this machine is made of unobtainium, as usual?
<maz> I could try and convince $WORK to get a handful if they were available.
<robmur01> quite a few RK3588 boards on the horizon now, so I suspect silicon might be near the point of mass production
<robmur01> I suspect I might end up buying one because MMU-600, but I'll refuse to enjoy it
* robmur01 still hasn't finished the DT for the RK3288 box that got him into this whole mess years ago
stikonas has joined #linux-rockchip
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #linux-rockchip
<Tenkawa> I've got the Odroid version of the RK3588 on order once it gets released.., also waiting on a RK3568 based Odroid based board to get here with several drive interface options so its going to be interesting to see how well the kernel and dtb comes out initially
<Tenkawa> so both are oredered... just waiting for them
matthias_bgg has quit [Ping timeout: 256 seconds]
vagrantc has joined #linux-rockchip
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #linux-rockchip
Tenkawa has quit [Quit: Was I really ever here?]
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #linux-rockchip
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #linux-rockchip
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #linux-rockchip
stikonas has quit [Read error: Connection reset by peer]
<macc24> is doc/README.rockchip still right about mainline u-boot being unable to boot from emmc on rk3399?
<robmur01> my RK3399 board running mainline everything off eMMC for the last 4 years or so suggests no
<robmur01> it probably still has the U-Boot hack to limit the eMMC to a 50MHz mode though, dunno if that ever got fixed properly
<macc24> robmur01: okay, how did you get it to boot from emmc?
<robmur01> AFAIR I built a U-Boot image and wrote it to the eMMC :/
<macc24> ah
<robmur01> eMMC is basically just another SD card in this context
macromorgan has joined #linux-rockchip
<macc24> o/ macromorgan
<robmur01> although I do have a vague memory of having to set the "same-as-spl" boot order property to convince it to behave, even though it seemed like that was supposed to be the default anyway
<robmur01> yeah, my local tree just has one patch doing that, on top of v2021.10, but according to `sudo strings /dev/mmcblk2` I'm still actually running something based on v2020.01
Tenkawa has joined #linux-rockchip
lurchi_ is now known as lurchi__
<Jmabsd> robmur01: feel free to write this up somewhere
Tenkawa has quit [Quit: Was I really ever here?]
Tenkawa has joined #linux-rockchip
<Jmabsd> Oh remind me, which RK3588 SBC:s are in the production pipe: Radxa, Firefly. Who more?
<robmur01> Huh, write what up where? Does it need recording for posterity that I can't be bothered to update bootloaders once they work?
<Tenkawa> Jmabsd Rock5
<Jmabsd> tenkawa: That's by Radxa. More?
<Tenkawa> Jmabsd: oh yeah duh
<Tenkawa> I was referencing it "from" the Odroid dealer... that was throwing me off
<Tenkawa> Jmabsd: QuartzPro64
<Tenkawa> by Pine
<macc24> so i kinda have a rk3399
<macc24> device
<macc24> that i can't kick out of rockusb mode
<macc24> it unfortunately has something on its emmc so it isn't maskrom mode and i can't override it
<macc24> but writes with rkdeveloptool seem to succeed, and from http://opensource.rock-chips.com/wiki_Rockusb i know that i have write access after 0x2000
<macc24> is there any way i can rescue that device without shorting emmc to ground and without uart?
<vagrantc> soldered emmc?
<macc24> yes
<vagrantc> i played that game with a firefly-rk3399 board ... no fun.
<macc24> i can't short emmc clock to ground because the device has no documentation :|
<robmur01> yeah, that state can be a bugger to get out of... I think if you have a fully-packed RKFW image, "uf" in rockusb can still overwrite the SPL along with everything else
<macc24> "fully packed rkfw image" so update.img?
<robmur01> ISTR rescuing my RK3288 box by doing a recovery boot to re-flash stock Android on several occasions
<Jmabsd> tenkawa: Oh yes
<robmur01> if there's enough SPL left and you're particularly lucky, it might still look to load U-Boot main stage off the SD card, but there are several variables at play there
<macc24> robmur01: there's enough u-boot there to recognize that it's broken and to kick rk3399 into rockusb
<macc24> and it tries complaining about u-boot being broken but i've got way too slow uart adapter
<vagrantc> oh, yeah, the default speed for many rk3399 is 1500000... you definitely want a reasonably ok uart...
<macc24> hmm i wonder if i can use stm32 board as a uart adapter
<robmur01> (because you can take it off and plumb in 1.8V or whatever for your wacky ODROIDs and 96boards too)
jakllsch_ has joined #linux-rockchip
<robmur01> Yeah, I've heard of folks running USB-UART example code on a Nucleo like that - not sure what speeds they can go up to though
<macc24> i also have an arduino clone with ch340g somewhere but it's 5v
jakllsch has quit [Ping timeout: 252 seconds]
<robmur01> breadboard up a couple of BJTs and resistors as a level shifter?
<macc24> i'll tinker with stm32 option
<vagrantc> sometimes i use https://www.crowdsupply.com/pylo/muart ... burned some usb ports once
<robmur01> the other thing is to look carefully around the vicinity of the eMMC (both sides of the board) for two close-together test pads where one is grounded and the other at around 1.8V
<robmur01> if they've followed the reference schematic without being too clever, that should be the EMMCCLK escape hatch :)
<macc24> hmm
jakllsch_ is now known as jakllsch
<robmur01> (although looking at rk3399_box_ref specifically I see it specifies a few more too - from experience EMMC_D0 should be equally sufficient to break booting; I guess EMMC_CMD might be; EMMC_STRB probably isn't)
<macc24> fuck yeah, got the uart log
<robmur01> looks like it might just be the miniloader left alive there
<macc24> flashed trust.img and got this
<macc24> "U-Boot 2017.09 (Mar 31 2022 - 09:55:55 +0800)" saved it :3
<robmur01> heh, I liked the DTB header where U-Boot was supposed to be :)
<macc24> now to erase miniloader completely and kick it into maskrom mode
<macc24> "DevNo=1 Vid=0x2207,Pid=0x330c,LocationID=102 Maskrom" \o/
<vagrantc> i make a habit of just installing my bootloader to microSD and not touching eMMC boot on systems where it is hard to recover from...
<macc24> device successfully bricked
<robmur01> hooray!
<macc24> "U-Boot TPL 2022.04-g42a2d90c-dirty (Apr 13 2022 - 23:43:47)", hello :>
<robmur01> dirty is the best kind of build
<robmur01> (which reminds me I haven't updated my local git version in ages...)
<macc24> now why does it complain of missing dtb
<robmur01> you flashed u-boot.itb, not u-boot.bin, right?
<macc24> yep
<macc24> note that i'm trying to port a board
<robmur01> oh, yeah, I have a half-finished attempt at that lying around too, and getting all the configs and stuff right seems to be a whole extra dose of fiddly
<Tenkawa> yay.. my RK3568 with NVME interface will be here in a few days
<Tenkawa> this should be fun
dvergatal has quit [Ping timeout: 256 seconds]
<robmur01> for certain values of "fun", at least
dvergatal has joined #linux-rockchip
<robmur01> I believe NVMe really wants working MSIs, which is not proving to be one of the RK35xx chips' strong points
Tenkawa has quit [Quit: Was I really ever here?]
<macc24> and i'm in spl now :D
<macc24> master branch might be broken on rk3399
lurchi__ is now known as lurchi_